The Parallel Agent Protocol — the communication protocol for intelligences that work together.
August 17, 2026
The Parallel Agent Protocol (PAP) is to agent computation what HTTP is to documents. An HTTP request retrieves a resource; a PAP request invokes an agent — it targets an invokable agent resource, causes computation, and returns structured results with execution metadata: timing, confidence, status, errors. Its URI schemes address a different kind of thing: not a page at https://, but an agent capability at pap:// — and paps:// when the operation involves execution, secrets or identity-bound action.
Labsintelligence · lab1 of Labs describe · invoke · stream
A protocol, not a third interface. IPI and OPI are programming interfaces; PAP is a different abstraction level entirely. IPI tells intelligence how to act. OPI tells intelligence how to coordinate. PAP lets intelligences communicate. The three are planes of one architecture, never rungs of one ladder.
01What “parallel” names
Parallel does not mean concurrent. It names parallel intelligences — autonomous computational actors existing alongside one another, each invokable, each able to answer for itself. Whether two invocations actually run at the same time is an orchestration decision, and the published Octra definition already makes parallel-versus-sequential execution exactly that. The division is precise:
PAP enables parallel agents. OPI determines parallel execution.
So when an orchestration says run A and B together, send both to C, escalate on disagreement — that is OPI speaking. What actually carries invoke A, invoke B, stream B’s progress, return A’s result across the boundary between intelligences is PAP. And where the invoked work finally touches an interactive environment, an IPI sits on the far end: OPI governs the work. PAP moves the work. IPI performs the work. A spoken ask — a Speech‑to‑Orchestration Voice Command at one end, Speech‑to‑Action Voice Commands at the other — can traverse all three planes without the visitor ever knowing the seams.
02The protocol, in brief
request
pap_version · id · target · op · headers · body · auth — three ops: describe (what can this agent do), invoke (do it), stream (do it, showing your work).
response
status · result or a typed error (parse_error, auth_error, policy_denied, timeout…) · execution metadata — time, confidence, trust — riding the response itself, never a side channel.
Every invocation passes the same governance as a live tool call — a PAP caller is never a side door around policy. The full wire format, error registry and versioning rules are in the specification.
03Where it stands
pap/1.0 is published as a specification, and its reference implementation is live inside LabsAI Studio — the Studio is the protocol’s first speaker. When a session’s orchestration delegates research, the step travels as a pap/1.0 invocation to the Studio’s own agent endpoint and returns with execution metadata; describe answers with the agent’s capability schemas; stream carries framed partial results. Today that traffic runs within a single node. The network of agent endpoints it is written for — discovery, resolution, trust between agents that have never met — is the road ahead, and the protocol is deliberately transport-independent so that road does not require rewriting it.
Contributing authors: Maya E. Davis · Duránd F. Davis Jr.
Tell us what you think, join us
This research is published while the questions are still open, and the systems it describes are live. We would love for you to join us — and please share your thoughts at research@labsintelligence.ai.
Citation
Please cite this work as:
Davis, Maya E., and Davis, Duránd F., Jr., “Introducing PAP.” LabsAI Research Briefs, Labsintelligence — lab1 of Labs Companies, Inc., August 2026.
Or use the BibTeX citation:
@article{labsintelligence2026introducingpap,
author = {Davis, Maya E. and Davis, Duránd F., Jr.},
title = {Introducing PAP},
journal = {LabsAI Research Briefs},
publisher = {Labsintelligence, lab1 of Labs Companies, Inc.},
year = {2026},
month = {august},
url = {https://labsintelligence.ai/research/labsai/introducing-pap/},
}
